Research On On-demand Transcoding Mechanism Of Live Streaming Based On Cloud Computing

In recent years,with the continuous development of media cloud technology and the popularity of live video services,the research on the transcoding mechanism of live video streaming based on cloud computing is of great practical and academic significance.The transcoding scheme of live streaming is more demanding than the on-demand video,requiring lower delay and higher quality of service(QoS).The live streaming processing technology based on cloud computing has more advantages than traditional CDN and P2 P technology,including better resource utilization,lower delay and more flexible solutions.However,there are still some unnecessary cost in the existing live cloud transcoding scheme.The Docker technology in cloud computing has more advantages than traditional virtual machine technology.Therefore,considering the above problems,we have studied the transcoding technology of live video streaming based on Docker cloud computing,the prediction model of user access which affects the deployment of transcoding nodes and the resource scheduling strategy based on QoS in cloud transcoding.The main works are as follows:(1)Research on the structure design of live system and transcoding module based on Cloud Computing.The traditional cloud transcoding architecture is basically based on VM to process VOD.In that system,it has a large resource storage module to store huge amounts of video resources,then the video streaming will be block transcoding distributed.But in live environment,video streaming can only be generated with time and there is no large block of video streaming to be processed.Therefore,based on the characteristics of live broadcast technology and streaming distribution technology,this paper improves with the traditional VOD cloud transcoding system by Docker technology.The system realizes the on-demand transcoding of live video streaming,and reduces the transmission delay and cost overhead through the reusing mechanism of the transcoding channel.(2)By predicting the access of live broadcast platform,it will contribute to the deployment of transcoding channels.The full multi-resolution transcoding of the live programs that few people watch will cause the rental cost of cloud resources to increase and the resource utilization rate to decrease.Therefore,this paper introduces intelligent prediction algorithm after analyzing the shortcomings of traditional prediction models,and finds the support vector regression(SVR)prediction model can match the prediction problem of access volume well.After that,particle swarm optimization(PSO)is used to optimize the parameters of SVR,which improves the prediction accuracy and provides theoretical support for the transcoding channel deployment.(3)In view of the demand that how to meet the low packet loss rate and not increase the extra cost in live broadcast,a predictive resource scheduling strategy based on QoS aware is proposed.On the basis of analyzing the effect of queuing model on task scheduling,the scheduling strategy that meets the requirements of performance is realized by forecasting the required resources,combining with periodic and remedial resource scheduling strategies.Experiments show that the proposed method can reduce packet loss rate and guarantee network performance without additional cost overhead.